Computer vision combined with a smartphone helps see colours accurately

Researchers are developing an application based on AI algorithms that runs on regular smartphones and utilises accurate hyperspectral imaging in an uncomplicated way.

A computer vision technology designed by researchers can soon turn your smartphone into an extremely accurate camera. Together with Assistant Professor Arto Klami, doctoral students of computer science Mikko Toivonen and Chang Rajani have designed computer vision algorithms under the spexel.ai project, which can be used to determine colours from photographs taken with a smartphone to a very high degree of precision.

Accurate colour detection can help, for example, interior designers or those who work with paint. With the spexel.ai technique, they can take a picture of their subject and obtain accurate information on its colour shade from the application. The technique could also be useful for online shoppers looking for a product of a specific hue.

Spexel.ai’s technology is based on the combination of a smartphone, a peripheral device attached to the phone and an application that determines the colour in a cloud-based service with the help of computer vision algorithms.

“There are other devices available for colour detection, but they need to be pointed very close to the object being measured. Our solution makes it possible to determine the colour also from further away, and it can also be used in taking images of, say, roofs,” Chang Rajani says.

Hyperspectral images reveal more

Colour detection constitutes the first step taken in spexel.ai’s product development. Subsequently, the same technology allows the utilisation of extremely precise hyperspectral imaging, where images taken with the peripheral device are processed into hyperspectral images with computer vision algorithms. In other words, the technology is more advanced than in colour detection.

Hyperspectral images are different from regular photographs due to the fact that they reveal things unseen to the naked eye in the object photographed. The technique is not based on transillumination; rather, hyperspectral images interpret the wavelengths of light more accurately than regular photos.

“Normal photos use three colour channels, such as red, blue and green. In hyperspectral imaging, the light wavelength resolution is finer, comprising a hundred colour channels,” Klami explains.

“A simple three-colour camera is unable to distinguish the spectrum of, for example, chlorophyll. In a hyperspectral image taken outdoors, it’s easier to identify the bits with chlorophyll, that is, the areas with vegetation,” Toivonen says.

The story continues after the video
Allow cookies to display content

This content is retrieved from www.youtube.com. Marketing cookies must be allowed to view this content. You can change your cookie settings via the cookie management link. You can also find the same link at the bottom of the site. If the content does not appear automatically after accepting cookies, please refresh the page.

Change cookie settings

Many uses, expensive equipment

The technology involved in hyperspectral imaging is already in use, among other fields, in geographical remote sensing and estimating yield sizes in agriculture. Hyperspectral images can also be used to identify counterfeit art and pharmaceuticals.

The technique has a range of uses also in industrial quality control, but so far it has not really been made available to consumers.

“Our vision is to bring hyperspectral imaging to consumers. In this, colour detection is a natural starting point, as the technical solutions required are easier to implement,” says Ville Kurri, commercial lead for spexel.ai.

Most of the devices with hyperspectral imaging capacity currently available are expensive specialist equipment which create images independently from start to finish. In the researchers’ version, images are converted from regular photographs taken with a smartphone equipped with a peripheral device, and the heavy lifting, or the development of the hyperspectral images, is carried out in a cloud service.

“The phone peripheral is cheap and compatible with basically all smartphones. Consumers don't need to buy a separate device,” Klami says.

At the moment, the spexel.ai developers are investigating the needs of potential customers and carrying on with proof of concepts and development as well as negotiations with investors. The commercialisation of the invention has been initiated, with a patent pending. An international patient examination procedure has also been launched to determine the preconditions for obtaining an international patent. The researchers aim to make the peripheral device and application available to consumers by 2021.

With spexel.ai's solution, accurate colour detection is made easy. The method provides valuable information for consumers as well as professionals working with colours.

Arto Klami, Mikko Toivonen and Chang Rajani. Photo: Susan Heikkinen

spexel.ai website

More information:

The article was originally published on 29 October 2019, after which an amended version was published on 23 October 2020.

Problem

Equipment needed in hyperspectral imaging costs thousands of euros and is unavailable to average consumers. And yet, accurate spectral images could also benefit consumers in their everyday life, as well as many businesses. Consumers can benefit also from accurate colour detection without hyperspectral imaging.

Solution

Spexel.ai, a research project ongoing at the University of Helsinki, is aiming to produce an inexpensive optical peripheral device and an application for smartphones, with which accurate spectral images can be taken by anyone. To begin with, the technique will be introduced to consumers as a colour detection application that helps to determine the shade of objects accurately.

Key benefits

Colour detection is useful for people involved in colour design as well as for consumers. The long-term goal is to radically reduce the cost of hyperspectral imaging. Photos taken with a normal smartphone and an inexpensive peripheral will be converted into finished images by computer vision algorithms.

Business model

A patent is pending for the invention, and preliminary commercialisation measures have been taken. The goal is to have a spin-out company up and running by 2021, introducing consumer devices and the application to the market. 

Join us 

We welcome investors and partners who utilise colour detection or hyperspectral imaging. We are also happy to collaborate with parties that are employing a solution of their own, which could be combined with hyperspectral imaging.

Get in touch to find out more!

Meet us on Node by Slush platform!